Fantasy Portrait in WAN 2.1 - AI Video No More Stone-Faced Avatars! (Patreon)

Tutorial Video : https://youtu.be/iJ4ehmFr2jg

This video dives into Fantasy Portrait, a powerful new AI animation framework developed by Alibaba’s team, built on the WAN 2.1 image-to-video model. It allows you to drive facial expressions from any source video—human or animal—onto AI-generated characters, including 3D models, cartoon avatars, or hyper-realistic portraits. Whether you're creating talking avatars, animated animals, or non-human characters for films, games, or ads, this tool offers stable, high-quality facial motion transfer with support for head turns and complex expressions. We walk through how to set it up locally using ComfyUI, integrate the updated WAN Video Wrapper, and generate expressive AI videos using landmarks and context chunking for longer clips. This is perfect for content creators, AI artists, indie filmmakers, and developers working on animated storytelling, virtual avatars, or digital marketing. If you're looking to streamline character animation and reduce production time and costs, this tutorial shows you how to leverage cutting-edge AI to bring any character to life.

FantasyPortrait

Enhancing Multi-Character Portrait Animation with Expression-Augmented Diffusion Transformers

FantasyPortrait : https://fantasy-amap.github.io/fantasy-portrait/

Github : https://github.com/Fantasy-AMAP/fantasy-portrait

For ComfyUI:

Custom Node: https://github.com/kijai/ComfyUI-WanVideoWrapper

FantasyPortrait AI Model File: https://huggingface.co/Kijai/WanVideo_comfy/tree/main/FantasyPortrait

Attached example workflow modify for long length video: